Unsworth et al. (2015) - Attention & Working Memory Lab

570367
research-article2015
PSSXXX10.1177/0956797615570367Unsworth et al.Video-Game Experience and Cognitive Abilities
Psychological Science OnlineFirst, published on April 20, 2015 as doi:10.1177/0956797615570367
Research Article
Is Playing Video Games Related to
Cognitive Abilities?
Psychological Science
1­–16
© The Author(s) 2015
Reprints and permissions:
sagepub.com/journalsPermissions.nav
DOI: 10.1177/0956797615570367
pss.sagepub.com
Nash Unsworth1, Thomas S. Redick2, Brittany D. McMillan1,
David Z. Hambrick3, Michael J. Kane4, and
Randall W. Engle5
1
University of Oregon; 2Purdue University; 3Michigan State University; 4University of North Carolina at
Greensboro; and 5Georgia Institute of Technology
Abstract
The relations between video-game experience and cognitive abilities were examined in the current study. In two
experiments, subjects performed a number of working memory, fluid intelligence, and attention-control measures and
filled out a questionnaire about their video-game experience. In Experiment 1, an extreme-groups analysis indicated
that experienced video-game players outperformed nonplayers on several cognitive-ability measures. However, in
Experiments 1 and 2, when analyses examined the full range of subjects at both the task level and the latentconstruct level, nearly all of the relations between video-game experience and cognitive abilities were near zero.
These results cast doubt on recent claims that playing video games leads to enhanced cognitive abilities. Statistical
and methodological issues with prior studies of video-game experience are discussed along with recommendations
for future studies.
Keywords
cognitive ability, individual differences
Received 10/1/14; Revision accepted 1/9/15
In recent years, there has been an explosion of research
examining whether prior experience playing video
games—in particular, action or first-person shooter
games—is related to various cognitive skills and abilities.
This research has demonstrated that video-game players
outperform non-video-game players on tests of attention
control, visual-spatial abilities, working memory, and
executive control (Blacker & Curby, 2013; Cain, Landau,
& Shimamura, 2012; Castel, Pratt, & Drummond, 2005;
Chisholm & Kingstone, 2012; Colzato, van den
Wildenberg, Zmigrod, & Hommel, 2013; Green & Bavelier,
2003, 2006, 2007; McDermott, Bavelier, & Green, 2014;
Powers, Brooks, Aldrich, Palladino, & Alfieri, 2013).
Furthermore, a typical finding across studies is that videogame players are faster on a number of measures
­compared with non-video-game players. These findings
have led many researchers to conclude that playing video
games can enhance a variety of cognitive abilities
(Bavelier, Green, Pouget, & Schrater, 2012; Granic, Lobel,
& Engels, 2014; Green & Bavelier, 2012; Spence & Feng,
2010). This conclusion has been bolstered by experimental training studies in which subjects who played a game
for a certain amount of time subsequently demonstrated
increased performance on a number of measures compared with control subjects who did not play the game
(Glass, Maddox, & Love, 2013; Green & Bavelier, 2006,
2007; Powers et al., 2013).
Despite the impressive number of studies that have
found positive results, other studies have failed to find
differences between video-game players and non-videogame players on a number of measures. For example,
although some studies have suggested that there are differences on measures of selective attention (Green &
Bavelier, 2003, 2006, 2007), others have failed to replicate
these effects (Irons, Remington, & McLean, 2011; Murphy
Corresponding Author:
Nash Unsworth, Department of Psychology, University of Oregon,
Eugene, OR 97403
E-mail: [email protected]
Unsworth et al.
2
& Spencer, 2009; Wilms, Petersen, & Vangkilde, 2013).
Likewise, some studies have suggested that there are differences on working memory measures (Blacker & Curby,
2013; Colzato et al., 2013; McDermott et al., 2014),
whereas others have not (Boot, Kramer, Simons, Fabiani,
& Gratton, 2008; Hambrick, Oswald, Darowski, Rench, &
Brou, 2010). Thus, despite the promise of examining
video-game playing as a means to enhance cognitive
abilities, it is still unclear how robust these effects are and
whether possible confounds, such as subject recruitment
and demand characteristics, may be influencing the
results (Boot, Blakely, & Simons, 2011; Green, Strobach,
& Schubert, in press).
In addition, studies that have compared video-game
players and non-video-game players have a number of
methodological and statistical problems. For example,
many of these studies have relied on very small sample
sizes (in some cases, as few as 7 to 8 subjects per group).
Given recent meta-analytic results suggesting that many
of these effects are small (Powers et al., 2013), it is possible that discrepancies in the literature are due to the
use of small sample sizes, which not only drastically
reduce the power to find small effects, but also increase
the likelihood of Type 1 errors (Button et al., 2013; Ingre,
2013).
Furthermore, most of these studies have used extremegroups designs, in which frequent video-game players
are compared with non-video-game players. In a typical
study, subjects with significant video-game experience
(typically 5+ hr a week) are compared with non-videogame players (less than 1 hr of video-game play a week)
on a variety of tasks, with more intermediate, casual gamers omitted. Extreme-groups designs can be problematic
for a number of reasons (Preacher, Rucker, MacCallum, &
Nicewander, 2005). For example, when only the top and
bottom portions of the distribution are examined, a great
deal of information is lost, as the entire middle of the
distribution has been excluded. Additionally, although
extreme-groups designs are known to increase the ability
to detect an effect, these designs can also lead to an
increased likelihood of making a Type I error as a result
of overestimated effect sizes (Conway et al., 2005;
Preacher et al., 2005). Finally, in extreme-groups designs,
subjects within a particular group are treated as equal
when they are not. In the case of video-game experience,
it is likely that most of the subjects in the non-gameplayer group are similar in the number of hours of game
play (they all report less than 1 hr per week), but it is
unlikely that subjects in the experienced group are so
similar; rather, variability in the experienced group will
likely be considerable and positively skewed, with some
subjects gaming only 6 hr per week and others playing
20 or more (see Latham, Patston, & Tippett, 2013, for a
similar critique).
Although extreme-groups studies have their place,
particularly in early stages of empirical investigation
(indeed, we have published a number of extreme-groups
studies), it is always desirable to subsequently test the
effects found in extreme-groups studies with a full range
of subjects and, ideally, to examine the results at the
latent-construct level. Given that video-game experience
(hours per week of game play) is a continuous variable,
the relation between game playing and cognitive abilities
should be examined across all subjects rather than just
compared between the extremes. Furthermore, given
issues with unreliability of measures and idiosyncratic
task effects, it is desirable to examine the relation between
video-game experience and cognitive abilities with multiple measures of cognitive ability and with latent-variable techniques. Finally, multiple constructs should be
examined simultaneously to ensure that any relations
found are not due to an unmeasured third variable. For
example, the relation between video-game experience
and working memory could be due to shared variance
with attention control.
With these issues in mind, we examined the relation
between video-game experience and cognitive abilities
(attention control, working memory, and fluid intelligence) in two experiments using fairly large samples of
subjects and measures. If video-game experience is
related to cognitive abilities, then these relations should
be found not only when extreme groups are compared,
but also when the full range of subjects is examined and
the relations are examined at the latent level.
Experiment 1
To examine whether playing video games is related to
cognitive abilities, we reanalyzed data from Unsworth
and McMillan (2014). In that study, a large number of
subjects completed multiple cognitive-ability measures
and filled out a questionnaire on their video-game
experience.
Method
Subjects. The data analyzed in Experiment 1 include a
subset of data reported in Unsworth and McMillan (2014).
The video-game data have not been published previously. A total of 252 subjects were recruited from the
subject pool at the University of Oregon; the recruitment
materials did not mention video-game experience or
playing, as recommended by Boot et al. (2011). Recruited
participants received course credit for their participation.
Data from 11 subjects who failed to complete two or
more of the cognitive-ability tasks were dropped. The
video-game-experience questionnaire was administered
to 218 of the subjects, 20 of whom failed to complete the
Video-Game Experience and Cognitive Abilities3
questionnaire and were excluded from data analysis. The
remaining 198 subjects were between the ages of 18 and
35 (M = 19.49, SD = 1.75). Subjects were tested in groups
of 1 to 6 in laboratory sessions lasting approximately 2
hr. The testing took place over two full academic quarters; the stopping rule was to end data collection at the
end of the second quarter. During the attention-control
tasks, subjects were periodically presented with thought
probes asking them to classify their immediately preceding thoughts. The results for these probes were reported
in Unsworth and McMillan and are not discussed here.
Procedure. After providing informed consent, all subjects completed the following tasks (in the order listed):
operation span, symmetry span, reading span, Raven’s
Advanced Progressive Matrices, number series, letter sets,
sustained attention to response (SART), antisaccade,
arrow flankers, Stroop, and psychomotor vigilance. All
tasks were computerized. Following the tasks, subjects
filled out a battery of questionnaires, including the videogame-experience questionnaire.
Working memory tasks. Working memory was tested
with three span tasks. A recalled item was scored as correct if the item had the correct identity (or location, in the
case of the symmetry-span task) and was recalled in the
correct list position. The score for each task was the total
number of correct items across all trials. Higher scores
represented better performance.
Operation span. On each trial in this task, subjects
solved a series of math problems while trying to remember a set of unrelated letters (taken from the set F, H,
J, K, L, N, P, Q, R, S, T, Y). After subjects solved a math
problem, they were presented with a letter for 1 s. Immediately afterward, the next problem was presented and
solved, and then another letter was presented. After this
alternating sequence occurred three to seven times, subjects were asked to recall the presented letters in order
(i.e., list lengths ranged from 3 to 7). Three trials of each
list length were presented, for a total possible of 75 letters
correctly recalled. The order in which the different list
lengths were presented varied randomly. At recall, subjects reported the list as they remembered it by clicking
on the appropriate letters (see Unsworth, Heitz, Schrock,
& Engle, 2005, and Redick et al., 2012, for more details).
Subjects received three practice trials (list length = 2).
Symmetry span. On each trial in this task, subjects
performed a symmetry-judgment task while trying to
remember a sequence of locations in which red squares
were presented within a matrix. Subjects viewed an 8 ×
8 matrix with some squares filled in black, reported
whether the design was symmetrical about its vertical
axis (which it was half of the time), and then were presented with a 4 × 4 matrix with one of the cells filled in
red for 650 ms. They then performed the judgment task
again and were presented with another matrix with a red
square. After this alternating sequence occurred two to
five times, subjects were asked to report the sequence
of red-square locations in the preceding displays (i.e.,
list lengths ranged from 2 to 5), in the order in which
they had appeared, by clicking on the cells of an empty
matrix (see Unsworth, Redick, Heitz, Broadway, & Engle,
2009, for more details). There were three trials of each
list length, for a total possible of 42 locations correctly
recalled.
Reading span. On each trial of this task, subjects were
required to read sentences while trying to remember a
set of unrelated letters (the same sets from the operationspan task were used). Subjects read a sentence (e.g., “The
prosecutor’s case was lost because it was not based on
fact”), reported whether it made sense (which it did half
the time), and then were presented with a letter for 1 s.
Immediately afterward, the next sentence was presented
and judged, and then another letter was presented. After
this alternating sequence occurred three to seven times,
subjects were asked to recall the presented letters in
order (i.e., list lengths ranged from 3 to 7), by clicking on
the appropriate letters. Nonsense sentences were made
by simply changing one word in an otherwise normal
sentence (e.g., changing “case” to “dish” in the preceding
example; see Unsworth et al., 2009, for more details on
this task). There were three trials of each list length, for a
total possible of 75 letters correctly recalled.
Fluid intelligence tasks
Raven’s Advanced Progressive Matrices. This test is a
measure of abstract, inductive reasoning (Raven, Raven,
& Court, 1998). Thirty-six items are presented in ascending order of difficulty. Each item consists of a display of
3 × 3 matrices of geometric patterns, arranged according
to an unknown set of rules, with the bottom right pattern
missing. The task is to select, among eight alternatives,
the one that correctly completes the overall series of patterns. After completing two practice problems, subjects
had 10 min to complete the 18 odd-numbered items from
the test. A subject’s score was the total number of correct
solutions. Higher scores represented better performance.
Number series. In this task, subjects saw a series of
numbers, arranged according to an unstated rule, and
were required to induce what the next number in the
series should be (Thurstone, 1962). Subjects selected
their answer from five possible numbers that were presented. After working on five practice items, subjects had
4.5 min to complete 15 test items. A subject’s score was
4
the total number of items solved correctly. Higher scores
represented better performance.
Letter sets. In this task, subjects saw five sets of four
letters and were required to induce a rule that described
the composition and ordering of four of the five sets
(Ekstrom, French, Harman, & Dermen, 1976). Subjects
were then required to indicate the set that violated the
rule. After working on two example problems, subjects
had 5 min to complete 20 test items. A subject’s score was
the total number of items solved correctly. Higher scores
represented better performance.
Attention-control tasks
SART. Subjects completed a version of the SART
with semantic stimuli, adapted from McVay and Kane
(2009). The SART is a go/no-go task in which subjects
must respond quickly with a key press to all presented
stimuli except infrequent (11%) target stimuli. In this
version of the SART, each word stimulus was presented
in 18-point Courier New font for 300 ms and followed
by a 900-ms mask. The nontarget stimuli were members
of one category (animals), and the infrequent targets
were members of a different category (foods). The SART
had 470 trials, on 50 of which targets were presented.
The dependent variables were accuracy on no-go trials
(i.e., successful withholding of responses) and standard
deviation of response time (RT) on go trials. Higher
accuracy and lower standard deviation represented better performance.
Antisaccade. Subjects began each trial in this task by
staring at a fixation point that was on-screen for a variable interval (200–2,200 ms). A white equal sign (“=”), the
cue, was then flashed either to the left or to the right of
fixation (11.33° of visual angle) for 100 ms. This was followed by a 50-ms blank screen and a second appearance
of the flashed cue for 100 ms on-screen. After another
50-ms blank screen, the target stimulus (“B,” “P,” or “R”)
appeared on-screen (11.33° to the left or right of fixation) for 100 ms, followed by masking stimuli (“H” for 50
ms and then “8” until a response was given). All stimuli
were presented in 12-point Courier New font. The subjects’ task was to report the target letter by pressing “1”
(for “B”), “2” (for “P”), or “3” (for “R”) on the number key
pad as quickly and accurately as possible. In the prosaccade condition, the flashing cue and the target appeared
in the same location. In the antisaccade condition, they
appeared in opposite locations. Subjects received, in
order, 10 practice trials to learn the response mapping,
10 trials of the prosaccade condition, and 50 trials of the
antisaccade condition. The dependent variable was accuracy on the antisaccade trials. Higher scores represented
better performance.
Unsworth et al.
Arrow flankers. Each trial in this task began with a
fixation point that was presented for 400 ms. An arrow
was then displayed directly above the fixation point for
1,700 ms. The subjects’ task was to indicate the direction
in which the arrow was pointing (by pressing the “F” key
for left-pointing arrows and the “J” key for right-pointing
arrows) as quickly and accurately as possible. On 30 neutral trials, the arrow was flanked by two horizontal lines
on each side. On 30 congruent trials, the flankers were
arrows pointing in the same direction as the target arrow.
Finally, on 30 incongruent trials, the flankers were arrows
pointing in the direction opposite the direction in which
the target arrow was pointing. The three trial types were
randomly intermixed. The dependent variable was the
RT difference between incongruent and congruent trials.
Lower scores represented better performance.
Stroop. On each trial of this task, subjects were presented with a color word (“red,” “green,” or “blue”) presented in one of three font colors (red, green, or blue).
The task was to indicate the font color via a key press
(“1” = red, “2” = green, “3” = blue), as quickly and accurately as possible. After completing 15 trials of responsemapping practice and 6 trials of practice with the real
task, subjects received 135 test trials. Of these trials,
67% were congruent; that is, the word’s meaning and
font color matched (e.g., “red” printed in red). The other
33% of the trials were incongruent (e.g., “red” printed in
green). The dependent variable was the RT difference
between incongruent and congruent trials. Lower scores
represented better performance.
Psychomotor vigilance. The psychomotor vigilance
task (Dinges & Powell, 1985) was used as the primary
measure of sustained attention. A row of zeroes appeared
on-screen, and after a variable amount of time, the zeros
began to count up to indicate the passage of time, in
1-ms intervals from 0 ms. The subjects’ task was to press
the space bar as quickly as possible once the counting
began. After the space bar was pressed, the current count
remained on-screen for 1 s to provide feedback on RT
for that trial. Interstimulus intervals were randomly distributed and ranged from 1 to 10 s. The entire task lasted
for 10 min (roughly 75 trials). The dependent variable
was the average RT for the slowest 20% of trials (Dinges
& Powell, 1985). Lower scores represented better performance.
Video-game-experience questionnaire. The videogame-experience questionnaire asked subjects to indicate their expertise (scale from 1 to 7; results not discussed
here) and the number of hours per week that they had
played various types of video games over the last year
(never, 0+ to 1, 1+ to 3, 3+ to 5, 5+ to 10, or 10+;
Video-Game Experience and Cognitive Abilities5
Table 1. Differences Between Video-Game Players and Non-Video-Game Players on the Cognitive-Ability
Measures in Experiment 1
Measure
Operation span
Symmetry span
Reading span
Raven’s matrices
Letter sets
Number series
Antisaccade
Flankers (ms)
SART SD
SART accuracy
Stroop (ms)
PVT (ms)
WM
gF
AC
Players (n = 18)
58.22
33.39
48.39
9.33
9.94
10.00
.51
124.27
152.98
.54
177.79
566.87
.12
.40
.11
(9.40)
(5.44)
(13.32)
(3.58)
(2.39)
(2.89)
(.15)
(81.95)
(36.80)
(.17)
(100.38)
(133.83)
(.70)
(.93)
(.81)
Nonplayers (n = 29)
52.48
27.03
50.66
6.69
8.57
7.24
.44
115.27
180.64
.53
179.23
692.16
–.22
–.50
–.34
(12.99)
(6.64)
(13.74)
(2.74)
(2.21)
(1.84)
(.08)
(129.86)
(56.42)
(.16)
(122.28)
(268.21)
(.82)
(.54)
(.73)
t(46)
p
Cohen’s d
rpb
1.63
3.41
–0.56
2.86
1.99
4.00
2.12
0.26
–1.85
0.21
–0.04
–1.84
1.46
4.22
1.98
.11
.00
.58
.00
.05
.00
.04
.79
.07
.83
.97
.07
.15
.00
.05
0.51
1.05
–0.17
0.83
0.59
1.14
0.67
0.08
–0.58
0.08
0.01
0.60
0.44
1.18
0.58
.24
.45
–.08
.39
.28
.51
.30
.04
–.27
.03
–.01
–.26
.21
.53
.28
Note: Video-game players and nonplayers were identified on the basis of their experience playing first-person shooter
games. SART = sustained-attention-to-response task; PVT = psychomotor vigilance task; WM = working memory factor
composite; gF = fluid intelligence factor composite; AC = attention-control factor composite; rpb = point-biserial correlation.
B. Hubert-Wallander, C. S. Green, & D. Bavelier, personal
communication, October 12, 2011). The different types of
games were first-person shooter games (e.g., Halo, Call
of Duty), action games (e.g., God of War, Mario Kart),
real-time strategy games (e.g., Warcraft, Starcraft), turnbased and puzzle games (e.g., Sims, Solitaire), role-­
playing games (e.g., World of Warcraft, Final Fantasy),
and music games (e.g., Guitar Hero, Rock Band).
Results
Extreme-groups analyses. For our first set of analyses,
we examined whether video-game players outperformed
non-video-game players on the different tasks. In particular, we were interested in whether subjects who reported
playing first-person shooter games frequently outperformed non-video-game players; most previous research
has specifically examined differences between individuals who self-report playing first-person shooter games
several hours per week and individuals who report playing no games. For this comparison, video-game players
were identified as subjects who reported playing firstperson shooter games for more than 5 hr per week over
the last year. Non-video-game players were identified as
those who reported both not playing first-person shooter
games at all and playing the other types of games less
than 1 hr a week. These criteria match those used in
many prior similar studies (e.g., Green & Bavelier, 2007;
McDermott et al., 2014). In this experiment, the criteria
identified 18 video-game players and 29 non-video-game
players. Table 1 summarizes their performance on each
measure and shows that the groups differed significantly
on a number of them: Video-game players outperformed
nongamers on one measure of working memory (symmetry span), all the fluid intelligence measures, and some
of the attention-control measures. Furthermore, across
most measures, there was a trend for video-game players
to outperform non-video-game players.
We also computed factor composites for the three different abilities (working memory, fluid intelligence, and
attention control) to get a sense for whether broader
measures of abilities would show group differences. As
Table 1 shows, the two groups differed on both fluid
intelligence and attention control, but the difference was
not significant for working memory. Overall, these results
are broadly consistent with prior extreme-groups studies
suggesting that video-game players have superior cognitive abilities compared with nonplayers.
Correlations. Next, we examined the zero-order correlations between video-game playing experience
(hours per week) for each game type and the cognitiveability measures, using the full range of subjects (N =
198). In other words, rather than examining extreme
groups, we utilized the full data set to determine the
relations between video-game playing and cognitive
abilities. Table 2 shows the descriptive statistics for all of
the measures for all subjects. Table 3 presents the zeroorder correlations between the cognitive-ability measures and self-reported experience playing each type of
video game. For first-person shooter games, only 3 of
the correlations were statistically significant, and all
Unsworth et al.
6
Table 2. Descriptive Statistics and Reliability Estimates for All Measures in Experiment 1
M
SD
Range
56.67
30.30
51.62
8.02
9.82
8.61
.48
112.06
152.44
.53
155.18
631.56
11.41
6.75
13.65
2.99
2.76
2.45
.12
80.32
50.75
.17
99.35
280.91
10–75
3–42
0–75
1–15
3–16
1–14
.21–.88
–44.89–697.50
35.61–352.83
.16–.98
–29.77–536.57
322.42–2,887.13
.88
.92
.29
.79
.35
.56
1.32
1.18
.85
.98
.99
.87
Measure
Cognitive-ability measures
Operation span
Symmetry span
Reading span
Raven’s matrices
Letter sets
Number series
Antisaccade
Flankers (ms)
SART SD
SART accuracy
Stroop (ms)
PVT (ms)
Video-game experiencea
Shooter games
Action games
Real-time strategy games
Turn-based games
Role-playing games
Music games
Reliability
.77
.71
.82
.70
.64
.70
.76
.60
.92
.83
.58
.90
—
—
—
—
—
—
0–5
0–5
0–5
0–5
0–5
0–5
Note: SART = sustained-attention-to-response task; PVT = psychomotor vigilance task; Reliability
estimates are from the full data set from Unsworth and McMillan (2014).
a
Subjects’ responses as to how many hours a week they played each category of video games were
coded on a scale from 0 (never) to 5 (10+ hr).
were relatively weak. Thus, whereas the extreme-groups
analysis suggested a number of significant relations
between cognitive abilities and experience playing firstperson shooter games, the correlational analysis, relying
on the full range of subjects, suggested fewer and weaker
relations. Similarly, weak to null relations were found for
action video games (1 significant correlation), real-time
strategy games (1 significant correlation), turn-based
games (3 significant correlations), role-playing games (no
significant correlations), and music games (1 significant
correlation in the nonpredicted direction). In total, only 8
out of a possible 72 correlations between video-game
Table 3. Zero-Order Correlations Between the Cognitive-Ability Measures and Video-Game
Experience for Each Type of Game in Experiment 1
Game type
Measure
Operation span
Symmetry span
Reading span
Raven’s matrices
Letter sets
Number series
Antisaccade
Flankers
SART SD
SART accuracy
Stroop
PVT
Shooter
Action
Real-time strategy
Turn-based
Role-playing
Music
.05
.14*
–.06
.14*
.05
.18*
.05
.05
–.06
–.03
.04
–.04
.08
.11
–.08
.11
.04
.20*
.04
–.06
–.05
.02
.06
–.07
–.02
.07
–.01
.22*
–.02
.06
.11
.10
–.05
.08
.06
.01
.03
.05
.18*
.07
.16*
–.04
.06
–.12
–.15*
–.10
.02
–.07
.04
.01
–.02
.08
–.03
.07
–.04
.10
–.03
.03
–.04
.03
–.10
–.10
.00
–.04
.02
–.14*
.11
.07
–.02
.02
.01
.05
Note: SART = sustained-attention-to-response task; PVT = psychomotor vigilance task.
*p < .05.
Video-Game Experience and Cognitive Abilities7
.49
Ospan
.71
.59
Symspan
.64
.55
Rspan
.73
Raven
.67
LS
.62
NS
.69
Anti
.89
Flanker
.84
SART SD
.87
SART Acc
.90
Stroop
.86
PVT
WM
.77
.67
.52
.58
gF
.68
.62
.75
.55
–.33
–.39
AC
.36
–.32
–.37
Fig. 1. Confirmatory factor analysis model of working memory (WM), fluid intelligence
(gF), and attention control (AC) in Experiment 1. Paths connecting the latent variables
(circles) to each other represent the correlations between the constructs, the numbers
from the latent variables to the manifest variables (squares) represent the loadings of
the tasks onto the corresponding latent variables, and the numbers next to the manifest
variables represent error variance associated with each task. Ospan = operation span;
Symspan = symmetry span; Rspan = reading span; Raven = Raven’s Advanced Progressive Matrices; LS = letter sets; NS = number series; Anti = antisaccade; Flanker = arrow
flankers; SART SD = standard deviation on the sustained-attention-to-response task;
SART Acc = accuracy on the sustained-attention-to-response task; PVT = psychomotor
vigilance task. All loadings and paths were significant, p < .05.
experience and cognitive abilities were statistically significant (p < .05) in the expected direction, and none of
the significant correlations were greater than .22 (53 correlations, or 74%, had absolute values < .10). We also
calculated the correlations separately for each gender to
see if gender had any influence and found that the correlations were virtually the same for males and females.
Latent-variable analyses. Although all of the cognitive-ability measures had adequate reliabilities (Table 2),
perhaps the weak to nonexistent relationship between
video-game experience and cognitive abilities was due to
idiosyncratic task effects. Examining these relations at the
latent-variable level, rather than the task level, should provide a better estimate of the relations between videogame experience and cognitive abilities. For this analysis,
we first specified a model for the cognitive-ability
measures using confirmatory factor analysis. Specifically,
we specified a model with a working memory latent variable (composed of the three working memory measures),
a fluid intelligence latent variable (composed of the three
fluid intelligence measures), and an attention-control
latent variable (composed of the six attention-control
measures). All measures were specified to load only on
the factor of interest, and all three latent variables were
free to correlate. The fit of the model was excellent,
χ2(51) = 46.99, p > .63, χ2/df = 0.92, root-mean-square
error of approximation (RMSEA) = .00, standardized rootmean-square residual (SRMR) = .04, nonnormed fit index
(NNFI) = 1.0, comparative fit index (CFI) = 1.0. As illustrated in Figure 1, each measure loaded significantly on its
factor of interest, and the factors were strongly intercorrelated, which is consistent with prior research (Unsworth,
Fukuda, Awh, & Vogel, 2014; Unsworth & Spillers, 2010).
Unsworth et al.
8
Table 4. Latent Correlations Between the Cognitive-Ability Measures and Video-Game Experience
for Each Type of Game in Experiment 1
Game type
Measure
Shooter
Working memory
Fluid intelligence
Attention control
.06
.23*
.04
Action
Real-time strategy Turn-based
.05
.23*
.08
.02
.14
.07
Role-playing
Music
.01
.08
–.04
–.10
–.12
.05
.13
.09
.12
*p < .05.
Next, we added in the video-game-experience measure for each type of video game. The video-game loadings were set equal to 1, and each type of video game
was allowed to correlate with each cognitive-ability factor and with the other video-game measures. The fit of
this model was also good, χ2(105) = 126.35, p > .08,
χ2/df = 1.20, RMSEA = .03, SRMR = .05, NNFI = .94, CFI =
.96. As Table 4 shows, the only significant latent correlations were between experience playing first-person
shooter games and fluid intelligence and between experience playing action games and fluid intelligence (both
rs = .23). All of the other relations were not significantly
different from zero, despite the large sample size. Thus,
as did the zero-order correlation analyses, the latent-­
variable analyses suggested that the relations between
video-game experience and cognitive abilities were weak
to nonexistent; only 2 out of a possible 18 correlations
were statistically significant (p < .05), which is not much
more than one would expect by chance.
trials in each attention-control task. RTs less than 200 ms
or 3 standard deviations above an individual’s mean were
excluded from all RT analyses. We then correlated RT in
each attention-control task with video-game experience
for each type of game. Table 5 shows the resulting zeroorder correlations. Once again, most of the correlations
were near zero; only 6 out of a possible 30 correlations
were significant, and these were weak. Furthermore, it
should be noted that 3 of these significant correlations (2
for the antisaccade task and 1 for the Stroop task) were
in the wrong direction, suggesting that more experience
playing video games was related with slower (not faster)
RTs. Thus, only 3 correlations actually suggested that
more video-game experience led to faster RTs.
To examine these issues further, we used confirmatory
factor analysis to specify a latent-variable model in which
a single RT latent variable was formed by having RTs
from the five attention-control tasks load on it; this factor
was allowed to correlate with experience playing each of
the different types of video games. The fit of this model
was acceptable, χ2(29) = 50.41, p < .01, χ2/df = 1.74,
RMSEA = .06, SRMR = .06, NNFI = .85, CFI = .90 and all
of the RT measures (except for the antisaccade task)
loaded significantly on the RT factor (antisaccade: .04;
flankers: .58; SART: .23; Stroop: .78; psychomotor vigilance: .34). The only latent-variable correlation between
video-game experience and RT that was significant was
the correlation for turn-based games, r = −.21. The other
RT correlations. Our final set of analyses examined
whether video-game experience was related to speed of
responding in attention-control tasks. Prior extremegroups studies (see Dye, Green, & Bavelier, 2009, for a
review) have consistently suggested that video-game
players are faster than non-video-game players on a variety of attention measures. To examine this relation, we
computed mean RT for correct responses only across all
Table 5. Zero-Order Correlations Between Mean Reaction Time on the Attention-Control Measures
and Video-Game Experience for Each Type of Game in Experiment 1
Game type
Measure
Antisaccade
Flankers
SART
Stroop
PVT
Shooter
Action
Real-time strategy
Turn-based
Role-playing
Music
–.01
–.04
–.05
.01
–.03
–.10
–.15*
.11
.16*
–.08
.16*
–.12
–.04
–.04
–.02
.06
–.16*
–.16*
–.11
–.09
.19*
–.08
–.02
–.03
–.01
.05
.04
–.04
.03
.01
Note: SART = sustained-attention-to-response task; PVT = psychomotor vigilance task.
*p < .05.
Video-Game Experience and Cognitive Abilities9
five correlations were nonsignificant—first-person shooter
games: r = −.03; action games: r = .08; real-time strategy
games: r = −.10; role-playing games: r = −.08; and music
games: r = .05. Thus, as in the prior analyses, these results
suggest weak to null correlations between video-game
experience and cognitive abilities (in this case, speed of
processing indexed by RT on the attention-control
measures).
Discussion
The extreme-groups analyses in Experiment 1 suggested
a number of significant effects indicating that video-game
players outperform non-video-game players on a variety
of cognitive measures. However, data for 151 subjects
had to be excluded from these analyses. In contrast,
when we calculated the zero-order and latent correlations using the full range of subjects, nearly all of these
effects were no longer significant and were very weak in
magnitude. The only consistent relations between videogame experience and cognitive abilities were the relations between fluid intelligence and experience with
first-person shooter and action games. However, these
effects should be interpreted cautiously, given that prior
extreme-groups and correlational studies have failed to
find a relation between action-video-game experience
and measures of fluid intelligence (Boot et al., 2008;
Colzato et al., 2013; Hambrick et al., 2010) and the metaanalytic effect size is rather small (Powers et al., 2013).
Thus, Experiment 1 suggests that the associations
between video-game playing and cognitive abilities are
weak to nonexistent.
There are two significant limitations to this experiment
that could hinder this interpretation. First, the video-game
questionnaire that we used (and that is used by many
researchers in the field) is categorical, and thus it is possible that we found only weak relations because there
was not enough variability between responses categories
to detect real relations. Second, although we relied on a
fairly large sample size, it is possible that there was range
restriction in the data; all of the subjects came from a
single university sample, and most of the students did not
report playing video games with high frequency.
Experiment 2
To rectify the limitations from Experiment 1, we reanalyzed data from Redick et al. (2014). The purpose of
Experiment 2 was to replicate and extend the results of
Experiment 1 using a larger and more representative
sample, a continuous (rather than categorical) measure
of video-game experience, and (again) a large number of
measures for each construct.
Method
Subjects. The data analyzed for this experiment are a
subset of the data reported in Redick et al. (2014). A total
of 586 subjects (226 male, 354 female, gender information missing for 6 subjects) were tested at four universities: Georgia Institute of Technology (Georgia Tech),
University of Georgia, Michigan State University, and
University of North Carolina at Greensboro. All subjects
were healthy young adults between the ages of 18 and
30. Not all were college students; 117 of the subjects
tested at the Georgia Tech site were community volunteers who were not attending that school (although some
attended other local universities). The recruitment materials made no mention of video-game experience. The
analyses reported here were based on only the 466 subjects who both completed all cognitive-ability tests in all
three test sessions (110 subjects did not have complete
data) and filled out the video-game questionnaire with
plausible answers (10 subjects did not provide plausible
answers, as described later). Subjects at the Georgia Tech
site were compensated $30 for each session; subjects at
all other sites received course credit in exchange for their
participation.
Procedure. All subjects completed the tasks in three
sessions, each lasting approximately 2 hr. At the beginning of the first session, subjects provided demographic
information and completed a questionnaire about their
video-game experience (Hambrick et al., 2010). The
demographic questions asked subjects their age, gender,
handedness, native language, and race. No information
was obtained on socioeconomic status, although future
research should address whether this variable influences
the relations between cognitive abilities and video-game
experience. All subjects completed the tasks in the same
order. The tasks in Session 1 were operation span, control tower, change detection, paper folding, arrow flankers, continuous counters, spatial Stroop, and reading
span. The tasks in Session 2 were matrix monitoring,
visual brief report, letter sets, cued visual search, keeping
track, symmetry span, SART, Raven’s, and Air Traffic Control (ATC) lab. Finally, in Session 3, subjects completed
the antisaccade, number-series, rotation-span, SynWin,
spatial delayed match-to-sample (DMTS), dual n-back,
cued-flankers, and math-access tasks. Data from the control-tower, ATC-lab, SynWin, spatial DMTS, dual n-back,
and math-access tasks were not analyzed in the current
study. All tasks were computerized.
Working memory tasks. The operation-span, symmetry-­
span, and reading-span tasks were the same as in Experiment 1.
10
Rotation span. In this task, subjects viewed rotated
letters and judged whether they were mirror-reversed
while they tried to remember a series of short or long
arrows each of which pointed in one of eight specific
directions. After two to five alternations of letters and
arrows, subjects were required to report the directions of
the arrows in serial order, choosing each in turn from the
full complement of possibilities. The score was the total
number of arrows recalled in the correct order (maximum = 42).
Keeping track. Each trial of this task began with the
presentation of exemplars from two to six different categories (tools, animals, etc.). Then, a list of 16 words was
presented, with each word shown on-screen for 1,500
ms (250-ms blank screen between words). After the final
word in the list was presented, six exemplars were shown
for each category, and subjects were asked to use the
mouse to click on the most recently presented exemplar
in each case. The total number of correctly recalled final
exemplars across the 15 lists was used as the dependent
variable. Higher scores indicated better performance.
Matrix monitoring. In each trial of this task, subjects
were presented with either one or two 4 × 4 matrices in
which a single cell was highlighted for 2,500 ms. Next,
a series of one to four arrows was presented, indicating
movement of the highlighted target cell or cells. A probe
matrix was displayed, and subjects had to report whether
the target cell would be in the indicated location in the
probe matrix if the target cell had moved as indicated by
the arrows. A single target matrix was presented on half
of the 16 trials, and two target matrices were presented
simultaneously on the other half. The proportion of correct responses to the probes was used as the dependent
variable. Higher scores indicated better performance.
Continuous counters. In this task, subjects were
instructed to keep separate running counts of the numbers of squares, circles, and triangles presented across
15 trials. Shapes were presented individually. Counting
the shapes of each type was difficult because the type of
shape changed randomly from trial to trial and the number of stimuli on each trial was unpredictable. The proportion of correct final counts was used as the dependent
variable. Higher scores indicated better performance.
Change detection. On each trial of this task, subjects
were briefly presented with a display of four to eight
colored squares. The display disappeared for 900 ms
and then reappeared with all the squares in the same
locations. One of the squares was circled, and subjects
reported whether that square was the same color it had
been when originally presented. There were 60 trials in
Unsworth et al.
all. A bias-corrected measure of capacity (k; Cowan et al.,
2005) was used as the dependent variable. Higher scores
indicated better performance.
Visual brief report. On each trial of this task, subjects
were presented with three to eight letters in random locations within a 4 × 4 grid. The letters were presented for
100 ms, and subjects then were instructed to report as
many letters as possible, by clicking on the appropriate letters. The total number of letters correctly reported
across 24 arrays was used as the dependent variable
(maximum = 132). Higher scores indicated better performance.
Fluid intelligence tasks. The Raven’s Advanced Progressive Matrices, letter-sets task, and number-series task
were the same as in Experiment 1.
For each problem of the paper-folding task, subjects
were presented with an illustration of a square piece of
paper on the left. Markings indicated how the paper had
been folded before a hole was punched through it. The
task was to decide which one of five response options
represented what the piece of paper would look like if it
was completely unfolded after the hole was punched.
The number of problems (out of 10) solved within the
4-min time limit was used as the dependent variable.
Higher scores indicated better performance.
Attention-control tasks
Antisaccade. This task was the same as in Experiment
1 except that the proportion of errors across a total of 30
trials was used as the dependent variable. Lower scores
indicated better performance.
Arrow flankers. This task was the same as in Experiment 1 except that there were 50 trials of each type.
SART. This task was the same as in Experiment 1
except that go stimuli occurred on 88.9% of all trials (540
trials total) and sensitivity (d′) and RT variability (standard deviation) for correct responses were the dependent variables.
Spatial Stroop. In this task, subjects were presented
with arrows and were asked to indicate whether each
arrow pointed to the left or right, while ignoring its location (left, center, or right portion of the computer screen).
These decisions were made difficult because on the
majority of the 144 trials, the location and direction information were congruent (e.g., a left-pointing arrow on the
left side of the screen), whereas on a small number of
trials, the location and direction information were incongruent (e.g., a left-pointing arrow on the right side of the
screen). The dependent variable was the RT ­difference
Video-Game Experience and Cognitive Abilities11
between incongruent and congruent trials. Lower scores
indicated better performance.
Cued visual search (Poole & Kane, 2009). On each
trial of this task, subjects decided whether an F located
within a 5 × 5 array of 25 letters (Es, backward Es, Fs,
backward Fs, 90°-tilted Ts, and 270°-tilted Ts) was mirror-reversed (facing left) or normal (facing right). Before
each multiletter visual array, subjects were given a 500ms arrow cue indicating in which 2 or 4 of the 25 possible array locations the F might appear (always along the
internal 3 × 3 “ring” of the array). Because other Fs were
randomly presented in noncued locations as irrelevant
distractors, subjects had to maintain the cue information
in order to respond correctly. The mean RT for correct
responses across the 160 trials was used as the dependent variable. Lower scores indicated better performance.
Cued flankers. On each trial in this task, subjects
decided whether a target F located within a horizontal
array of seven letters (including a distractor F) was mirrorreversed (facing left) or normal (facing right). Before each
array, subjects were given a location cue (a digit presented
for 500 ms) that indicated which of the following seven
letters was the target. The target was restricted to one of
the middle five positions in the row (i.e., Locations 2–6).
There were 120 trials in total. On compatible trials, all of
the letters faced the same direction as the target. However,
on the 20 incompatible-lure trials, the target and some of
the irrelevant flanker letters faced in opposite directions,
and subjects had to remember the cue information to
respond correctly. The mean RT for correct responses on
the incompatible-lure trials was used as the dependent
variable. Lower scores indicated better performance.
Video-game-experience questionnaire. The videogame-experience questionnaire asked subjects to estimate the number of hours per week that they played
different types of video games (Hambrick et al., 2010).
Before providing these estimates, subjects were explicitly
told that “with 24 hours per day, and 7 days per week, the
maximum number of hours is 168.” The different types of
games were first- and third-person shooter games (e.g.,
Halo, Medal of Honor, Call of Duty, Gears of War, Resident Evil); sports, action, and “other” games (e.g., Madden, Wii Sports, Super Mario Brothers, Sonic, Tetris);
real-time strategy games (e.g., Age of Empires, Starcraft,
Halo Wars); and role-playing games (e.g., World of Warcraft, Grand Theft Auto, Fable, Final Fantasy). Of the 10
subjects excluded on the basis of their responses to this
questionnaire, 9 provided total values that equaled or
exceeded 168 hr per week; 1 additional subject reported
playing first- and third-person shooter games for 121 hr
each week.
Results
Correlations. First, we examined the zero-order correlations between video-game-playing experience (hours
per week) and the cognitive-ability measures, using the
full range of subjects. Table 6 shows the descriptive statistics for all measures. On average, subjects played video
games for a total of 6.5 hr per week (SD = 14.67) and
played each individual type of game roughly 1 to 2 hr a
week (Fig. S1 in the Supplemental Material shows the
distribution of the number of hours of playing time for
each type of video game). Table 7 presents the zeroorder correlations between the cognitive-ability measures
and responses on the video-game questionnaire for the
four types of video games. Only 4 of the correlations
were statistically significant (1 was in the nonpredicted
direction), and all were relatively weak. Thus, only 3 out
of a possible 80 correlations between video-game experience and cognitive abilities were significant in the
expected direction, and none were greater than .13. Note
that for the RT measures, only 1 correlation was significant (that between cued visual search and experience
with role-playing games). Thus, the data provide little
evidence for a relation between processing speed and
video-game experience. As in Experiment 1, we also
examined the correlations separately for males and
females and found that there were no differences as a
function of gender.
Latent-variable analyses. As in Experiment 1, we
next used latent-variable techniques to better examine
the relation between video-game experience and cognitive abilities. We first specified a model for the cognitive-ability measures using confirmatory factor analysis.
As before, the model had a working memory latent
variable (composed of the nine working memory measures), a fluid intelligence latent variable (composed of
the four fluid intelligence measures), and an attentioncontrol latent variable (composed of the seven attention control measures). All measures were specified to
load only on the factor of interest, and all three latent
variables were free to correlate. The fit of the model
was good, χ2(167) = 658.71, p < .01, χ2/df = 3.94,
RMSEA = .08, SRMR = .06, NNFI = .93, CFI = .94. As
shown in Figure 2, each measure loaded significantly
on its factor of interest, and the factors were strongly
intercorrelated.
Next, we added in the video-game-experience measures. The video-game loadings were set equal to 1, and
each type of video game was allowed to correlate with
each cognitive-ability factor and with the other videogame measures. The fit of this model was also adequate,
χ2(235) = 738.09, p < .01, χ2/df = 3.14, RMSEA = .07,
SRMR = .06, NNFI = .93, CFI = .94. As Table 8 shows,
Unsworth et al.
12
Table 6. Descriptive Statistics and Reliability Estimates for All Measures in Experiment 2
Measure
M
Cognitive-ability measures
Operation span
54.75
Reading span
52.30
Symmetry span
25.94
Rotation span
27.82
Change detection (k)
3.44
Continuous counters
.83
Keeping track
35.14
Brief report
94.46
Matrix monitoring
.81
Antisaccade
.47
Arrow flankers (ms)
100.15
Cued flankers (ms)
717.34
Cued visual search (ms)
1,258.58
1.69
SART d′
SART SD
162.90
Spatial Stroop (ms)
137.12
Letter sets
10.33
Raven’s matrices
9.14
Number series
8.83
Paper folding
5.94
Video-game experience (hr/week)
Shooter games
1.70
Strategy games
0.76
Role-playing games
1.93
Sports, action games
2.14
SD
Range
14.45
14.11
8.88
8.69
1.27
.18
8.82
11.00
.14
.16
67.94
245.52
277.50
1.05
49.07
74.20
3.13
3.57
2.96
2.65
2–75
0–75
0–42
0–42
–1.33–5.80
.16–1.00
6–53
64–119
.31–1.00
.07–.80
–276.94–518.64
210.00–1,824.93
519.69–2,468.99
–3.61–3.90
50.35–354.34
–7.09–487.33
1–18
0–17
1–15
0–10
5.58
3.57
6.80
5.30
0–48
0–45
0–78
0–40
Reliability
.84
.83
.81
.80
.78
.91
.85
.73
.57
.86
.61
.87
.89
.96
.86
.48
.72
.79
.75
.78
—
—
—
—
Note: SART = sustained-attention-to-response task. Reliability estimates are from the full data set in Redick
et al. (2014).
none of the 12 latent correlations between video-game
experience and cognitive abilities was statistically significant. Thus, as did the zero-order correlation analyses, the
latent-variable analyses suggested that the relations
between video-game experience and cognitive abilities
were weak to nonexistent.
Discussion
Using a large and more representative sample of subjects, along with a continuous measure of video-game
experience, Experiment 2 largely replicated Experiment
1 in finding weak to null correlations between videogame experience and measures of cognitive abilities.
Experiment 2 also extended the results from Experiment
1 to a number of additional measures of working memory and attention control. Finally, unlike in Experiment
1, the relation between video-game experience and
fluid intelligence was not significant in Experiment 2.
These results, coupled with prior research that failed to
find a relation between video-game experience and
fluid intelligence (Boot et al., 2008; Colzato et al., 2013;
Hambrick et al., 2010), cast doubt on the stability of this
relation.
General Discussion
In the current study, we examined whether experience
playing video games is related to cognitive abilities using
large samples of subjects and a variety of tasks. In
Experiment 1, we examined the data using an extremegroups design, as has been done previously in the literature, and found a number of significant effects suggesting
that video-game players outperformed non-video-game
players on a variety of cognitive measures. However,
when we analyzed the data in Experiments 1 and 2 using
the full range of subjects, many of these effects were no
longer significant and were very weak in magnitude.
Likewise, when we examined the relationships using
latent variables, which more accurately represent the
intended constructs, most of the latent correlations in
Experiments 1 and 2 were near zero; those that were
significant were very weak compared with the relationships among the cognitive-abilities factors themselves.
Video-Game Experience and Cognitive Abilities13
Table 7. Zero-Order Correlations Between the Cognitive-Ability Measures and VideoGame Experience for Each Type of Game in Experiment 2
Game type
Measure
Operation span
Reading span
Symmetry span
Rotation span
Change detection
Continuous counters
Keeping track
Brief report
Matrix monitoring
Antisaccade
Arrow flankers
Cued flankers
Cued visual search
SART d′
SART SD
Spatial Stroop
Letter sets
Raven’s matrices
Number series
Paper folding
Shooter
Strategy
Role-playing
Sports, action
.01
–.03
.02
.04
.02
–.08
–.13*
.06
–.04
–.05
–.06
–.03
–.03
–.05
.03
–.08
–.01
–.05
.00
–.01
.00
.05
.03
.08
.05
–.07
–.06
.03
.00
–.09*
–.06
–.02
–.05
–.05
.05
–.03
.03
–.02
–.03
–.01
–.02
–.04
.05
.06
–.01
–.04
–.07
.04
.00
–.07
–.04
–.08
–.10*
–.06
.01
–.05
.00
.00
.02
.03
.03
–.04
–.03
.05
–.02
–.07
–.05
.13*
–.02
–.03
–.06
–.08
–.05
–.01
–.03
.03
–.02
–.06
.00
–.03
Note: SART = sustained-attention-to-response task.
*p < .05.
That is, the presence of strong intercorrelations among
working memory, fluid intelligence, and attention control
indicates that there was sufficient variability present
among the subjects to allow for strong relationships with
the video-game variables to be observed, if they existed.
These results are in direct contrast with those of many
prior studies that have suggested a strong relationship
between video-game experience and cognitive abilities.
One potential reason for these discrepant findings is that
many prior studies relied on relatively small sample sizes
and extreme-groups designs, which resulted in overestimated effect sizes and increased likelihood of Type 1
errors. Indeed, by analyzing the data in Experiment 1
both ways, we demonstrated that effect sizes obtained
from extreme-groups analyses were much larger than
those obtained from analyses including the full range of
subjects. Specifically, as shown in Table 1, the point-biserial correlations for several of the significant relations
between experience playing first-person shooter games
and the cognitive-ability measures had values greater
than .30 in the extreme-group comparisons. However,
the correlations for the exact same relations were smaller
than .25 when calculations were based on the full range
of subjects. Thus, the effect sizes obtained using the full
range of subjects were drastically smaller than those
found when we relied on extreme-groups analyses,
which suggests that prior extreme-groups studies overestimated the effect sizes and may have found spurious
relations as a result. We also suggest that effect sizes were
likely overestimated in the prior meta-analysis, given that
it was largely based on extreme-groups studies (Powers
et al., 2013; see Redick & Webster, 2014, for a similar
critique).
Overall, the current results suggest weak to nonexistent relations between video-game experience—across
a variety of different games—and fundamental cognitive
abilities (working memory, fluid intelligence, attention
control, and speed of processing). In order to fruitfully
examine whether playing video games is related to cognitive abilities, future research should examine the full
range of subjects (i.e., not rely exclusively on extremegroups analyses), should rely on sufficiently large sample sizes to estimate stable correlational effects, and
should examine these relations across a number of similar measures thought to represent the constructs of
interest (ideally using latent-variable techniques). More
research is needed to examine whether there are any
stable relations between video-game experience and
cognitive abilities and, if there are, what their overall
magnitude is.
Unsworth et al.
14
.49
Ospan
.51
.59
Symspan
.66
.55
Rspan
.49
Rotspan
.68
.59
Array K
.60
.55
Counters
.62
WM
.76
.65
.49
Keep Track
.59
Brf Report
.55
Matrix Mon
.73
Raven
.67
LS
.62
NS
.40
.90
.61
.78
.67
.76
gF
.72
.55
.62
Paper
.69
Anti
.89
Flanker
.84
SART SD
.87
SART d ′
–.27
.90
Spa Stroop
–.50
.86
C Flanker
.86
C Search
.75
–.58
–.17
–.71
AC
.63
–.42
Fig. 2. Confirmatory factor analysis model of working memory (WM), fluid intelligence (gF),
and attention control (AC) in Experiment 2. Paths connecting the latent variables (circles) to
each other represent the correlations between the constructs, the numbers from the latent
variables to the manifest variables (squares) represent the loadings of the tasks onto the corresponding latent variables, and the numbers next to the manifest variables represent error variance associated with each task. Ospan = operation span; Symspan = symmetry span; Rspan =
reading span; Rotspan = rotation span; Array K = change detection; Counters = continuous
counters; Keep Track = keeping track; Brf Report = visual brief report; Matrix Mon = matrix
monitoring; Raven = Raven’s Advanced Progressive Matrices; LS = letter sets; NS = number
series; Paper = paper folding; Anti = antisaccade; Flanker = arrow flankers; SART SD = standard
deviation on the sustained-attention-to-response task; SART d′ = d′ on the sustained-attentionto-response task; Spa Stroop = spatial Stroop; C Flanker = cued flankers; C Search = cued visual
search. All loadings and paths were significant, p < .05.
Video-Game Experience and Cognitive Abilities15
Table 8. Latent Correlations Between the Cognitive-Ability
Measures and Video-Game Experience for Each Type of Game
in Experiment 2
Game type
Measure
Working memory
Fluid intelligence
Attention control
Shooter Strategy Role-playing
–.04
–.03
.01
.01
–.02
.01
–.01
.01
.04
Sports,
action
–.03
–.04
.05
*p < .05.
Author Contributions
N. Unsworth developed the study concept. B. D. McMillan and
N. Unsworth designed Experiment 1 and conducted the testing
and data collection. R. W. Engle, D. Z. Hambrick, M. J. Kane,
T. S. Redick, and N. Unsworth designed Experiment 2 and conducted the testing and data collection. N. Unsworth performed
the data analysis. N. Unsworth drafted the manuscript, and T. S.
Redick, B. D. McMillan, D. Z. Hambrick, M. J. Kane, and R. W.
Engle provided critical revisions. All authors approved the final
version of the manuscript for submission.
Declaration of Conflicting Interests
The authors declared that they had no conflicts of interest with
respect to their authorship or the publication of this article.
Funding
Part of this work was supported by grants from the Office of
Naval Research (N00014-12-1-0406 and N00014-12-1-1011) to
R. W. Engle.
Supplemental Material
Additional supporting information can be found at http://pss
.sagepub.com/content/by/supplemental-data
References
Bavelier, D., Green, C. S., Pouget, A., & Schrater, P. (2012).
Brain plasticity through the life span: Learning to learn and
action videogames. Annual Review of Neuroscience, 35,
391–416.
Blacker, K. J., & Curby, K. M. (2013). Enhanced visual shortterm memory in action videogame players. Attention,
Perception, & Psychophysics, 75, 1128–1136.
Boot, W. R., Blakely, D. P., & Simons, D. J. (2011). Do action
video games improve perception and cognition? Frontiers
in Psychology, 2, Article 226. Retrieved from http://journal
.frontiersin.org/article/10.3389/fpsyg.2011.00226/full
Boot, W. R., Kramer, A. F., Simons, D. J., Fabiani, M., & Gratton,
G. (2008). The effects of video game playing on attention,
memory, and executive control. Acta Psychologica, 129,
387–398.
Button, K. S., Ioannidis, J. P. A., Mokrysz, C., Nosek, B. A.,
Flint, J., Robinson, E. S. J., & Munafò, M. R. (2013). Power
failure: Why small sample size undermines the reliability of
neuroscience. Nature Reviews Neuroscience, 14, 365–376.
Cain, M. S., Landau, A. N., & Shimamura, A. P. (2012). Action
video game experience reduces the cost of switching tasks.
Attention, Perception, & Psychophysics, 74, 641–647.
Castel, A. D., Pratt, J., & Drummond, E. (2005). The effects of
action video game experience on the time course of inhibition of return and the efficiency of visual search. Acta
Psychologica, 119, 217–230.
Chisholm, J. D., & Kingstone, A. (2012). Improved top-down
control reduces oculomotor capture: The case of action
video game players. Attention, Perception, & Psychophysics,
74, 257–262.
Colzato, L. S., van den Wildenberg, W. P. M., Zmigrod, S., &
Hommel, B. (2013). Action video gaming and cognitive
control: Playing first person shooter games is associated
with improvement in working memory but not action inhibition. Psychological Research, 77, 234–239.
Conway, A. R. A., Kane, M. J., Bunting, M. F., Hambrick, D. Z.,
Wilhelm, O., & Engle, R. W. (2005). Working memory
span tasks: A methodological review and user’s guide.
Psychonomic Bulletin & Review, 12, 769–786.
Cowan, N., Elliott, E. M., Saults, J. S., Morey, C. C., Mattox,
S., Hismjatullina, A., & Conway, A. R. A. (2005). On the
capacity of attention: Its estimation and its role in working
memory and cognitive aptitudes. Cognitive Psychology, 51,
42–100.
Dinges, D. F., & Powell, J. W. (1985). Microcomputer analyses of performance on a portable, simple visual RT task
during sustained operations. Behavior Research Methods,
Instruments, & Computers, 17, 652–655.
Dye, M. W. G., Green, C. S., & Bavelier, D. (2009). Increasing
speed of processing with action video games. Current
Directions in Psychological Science, 18, 321–326.
Ekstrom, R. B., French, J. W., Harman, H. H., & Dermen, D.
(1976). Manual for kit of factor-referenced cognitive tests.
Princeton, NJ: Educational Testing Service.
Glass, B. D., Maddox, W. T., & Love, B. C. (2013). Real-time
strategy game training: Emergence of a cognitive flexibility trait. PLoS ONE, 8(8), Article e70350. Retrieved from
http://journals.plos.org/plosone/article?id=10.1371/journal
.pone.0070350
Granic, I., Lobel, A., & Engels, R. C. M. E. (2014). The benefits of playing video games. American Psychologist, 69,
66–78.
Green, C. S., & Bavelier, D. (2003). Action video game modifies
visual selective attention. Nature, 423, 534–537.
Green, C. S., & Bavelier, D. (2006). Effect of action video
games on the spatial distribution of visuospatial attention.
Journal of Experimental Psychology: Human Perception
and Performance, 32, 1465–1478.
Green, C. S., & Bavelier, D. (2007). Action-video-game experience alters the spatial resolution of vision. Psychological
Science, 18, 88–94.
Green, C. S., & Bavelier, D. (2012). Learning, attentional control,
and action video games. Current Biology, 22, R197–R206.
Green, C. S., Strobach, T., & Schubert, T. (in press). On methodological standards in training and transfer experiments.
Psychological Research.
16
Hambrick, D., Oswald, F., Darowski, E., Rench, T., & Brou, R.
(2010). Predictors of multitasking performance in a synthetic work paradigm. Applied Cognitive Psychology, 24,
1149–1167.
Ingre, M. (2013). Why small low-powered studies are worse
than large high-powered studies and how to protect against
“trivial” findings in research: Comment on Friston (2012).
NeuroImage, 81, 496–498.
Irons, J. L., Remington, R. W., & McLean, J. P. (2011). Not so fast:
Rethinking the effects of action video games on attentional
capacity. Australian Journal of Psychology, 63, 224–231.
Latham, A. J., Patston, L. L. M., & Tippett, L. J. (2013). Just
how expert are “expert” video-game players? Assessing
the experience and expertise of video-game players across
“action” video-game genres. Frontiers in Psychology, 4,
Article 941. Retrieved from http://journal.frontiersin.org/
article/10.3389/fpsyg.2013.00941/full
McDermott, A. F., Bavelier, D., & Green, C. S. (2014). Memory
abilities in action video game players. Computers in Human
Behavior, 34, 69–78.
McVay, J. C., & Kane, M. J. (2009). Conducting the train of thought:
Working memory capacity, goal neglect, and mind wandering in an executive-control task. Journal of Experimental
Psychology: Learning, Memory, and Cognition, 35, 196–204.
Murphy, K., & Spencer, A. (2009). Playing video games does not
make for better visual attention skills. Journal of Articles in
Support of the Null Hypothesis, 6, 1–20.
Poole, B. J., & Kane, M. J. (2009). Working-memory capacity predicts the executive control of visual search among
distractors: The influences of sustained and selective attention. The Quarterly Journal of Experimental Psychology, 62,
1430–1454.
Powers, K. L., Brooks, P. J., Aldrich, N. J., Palladino, M. A., &
Alfieri, L. (2013). Effects of video-game play on information
processing: A meta-analytic investigation. Psychonomic
Bulletin & Review, 20, 1055–1079.
Preacher, K. J., Rucker, D. D., MacCallum, R. C., & Nicewander,
W. A. (2005). Use of the extreme groups approach: A critical
reexamination and new recommendations. Psychological
Methods, 10, 178–192.
Raven, J. C., Raven, J. E., & Court, J. H. (1998). Progressive
Matrices. Oxford, England: Oxford Psychologists Press.
Unsworth et al.
Redick, T. S., Broadway, J. M., Meier, M. E., Kuriakose, P. S.,
Unsworth, N., Kane, M. J., & Engle, R. W. (2012). Measuring
working memory capacity with automated complex span
tasks. European Journal of Psychological Assessment, 28,
164–171.
Redick, T. S., Shipstead, Z., Meier, M. E., Evans, J., Hicks, K. L.,
Unsworth, N., . . . Engle, R. W. (2014). Cognitive predictors
of a common multitasking ability. Manuscript submitted for
publication.
Redick, T. S., & Webster, S. B. (2014). Videogame interventions and spatial ability interactions. Frontiers in Human
Neuroscience, 8, Article 183. Retrieved from http://journal
.frontiersin.org/article/10.3389/fnhum.2014.00183/full
Spence, I., & Feng, J. (2010). Video games and spatial cognition. Review of General Psychology, 14, 92–104.
Thurstone, T. G. (1962). Primary mental abilities. Chicago, IL:
Science Research Associates.
Unsworth, N., Fukuda, K., Awh, E., & Vogel, E. K. (2014).
Working memory and fluid intelligence: Capacity, attention
control, and secondary memory. Cognitive Psychology, 71,
1–26.
Unsworth, N., Heitz, R. P., Schrock, J. C., & Engle, R. W. (2005).
An automated version of the operation span task. Behavior
Research Methods, 37, 498–505.
Unsworth, N., & McMillan, B. D. (2014). Similarities and differences between mind-wandering and external distraction: A latent variable analysis of lapses of attention and
their relation to cognitive abilities. Acta Psychologica,
150, 14–25.
Unsworth, N., Redick, T. S., Heitz, R. P., Broadway, J., & Engle,
R. W. (2009). Complex working memory span tasks and
higher-order cognition: A latent variable analysis of the
relationship between processing and storage. Memory, 17,
635–654.
Unsworth, N., & Spillers, G. J. (2010). Working memory capacity: Attention, memory, or both? A direct test of the dualcomponent model. Journal of Memory and Language, 62,
392–406.
Wilms, I. L., Petersen, A., & Vangkilde, S. (2013). Intensive
video gaming improves encoding speed to visual shortterm memory in young male adults. Acta Psychologica,
142, 108–118.